22
PHYLOGENY AND PHYLOGENETIC TAXONOMY OF DIPSACALES, WITH SPECIAL REFERENCE TO SINADOXA AND TETRADOXA (ADOXACEAE) MICHAEL J. DONOGHUE, 1 TORSTEN ERIKSSON, 2 PATRICK A. REEVES, 3 AND RICHARD G. OLMSTEAD 3 Abstract. To further clarify phylogenetic relationships within Dipsacales,we analyzed new and previously pub- lished rbcL sequences, alone and in combination with morphological data. We also examined relationships within Adoxaceae using rbcL and nuclear ribosomal internal transcribed spacer (ITS) sequences. We conclude from these analyses that Dipsacales comprise two major lineages:Adoxaceae and Caprifoliaceae (sensu Judd et al.,1994), which both contain elements of traditional Caprifoliaceae. Within Adoxaceae, the following relation- ships are strongly supported: (Viburnum (Sambucus (Sinadoxa (Tetradoxa, Adoxa)))). Combined analyses of C ap ri foliaceae yield the following: (Caprifolieae (Diervilleae (Linnaeeae (Morinaceae (Dipsacaceae (Triplostegia, Valerianaceae)))))). On the basis of these results we provide phylogenetic definitions for the names of several major clades. Within Adoxaceae, Adoxina refers to the clade including Sinadoxa, Tetradoxa, and Adoxa. This lineage is marked by herbaceous habit, reduction in the number of perianth parts,nectaries of mul- ticellular hairs on the perianth,and bifid stamens. The clade including Morinaceae, Valerianaceae, Triplostegia, and Dipsacaceae is here named Valerina. Probable synapomorphies include herbaceousness,presence of an epi- calyx (lost or modified in Valerianaceae), reduced endosperm,and distinctive chemistry,including production of monoterpenoids. The clade containing Valerina plus Linnaeeae we name Linnina. This lineage is distinguished by reduction to four (or fewer) stamens, by abortion of two of the three carpels,and possibly by supernumerary inflorescences bracts. Keywords: Adoxaceae, Caprifoliaceae, Dipsacales, ITS, morphological characters, phylogeny, phylogenetic taxonomy, phylogenetic nomenclature, rbcL, Sinadoxa, Tetradoxa. We are grateful to Jim Smith for rbcL sequencing efforts early in the life of this project, and to Bruce Baldwin for help with ITS sequencing. We also thank D. Boufford, B. Tan, K. In (Chengdu, Sichuan), T.-S.Ying (Beijing), and T.-N. Ho (Xining, Qinghai) for organizing expeditions to re-collect Tetradoxa and Sinadoxa, and for their help in the field. The expedition to Qinghai was supported by a National Geographic Society award to D. Boufford. Nancy Pyck and Chuck Bell shared information on their ndhF studies of Dipsacales, and Anders Backlund provided useful comments early on. 1 Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut, 06511-8160 U.S.A. E-mail: [email protected] 2 Bergius Foundation, Royal Swedish Academy of Sciences, Box 50017, S-104 05 Stockholm, Sweden,and Department of Botany, Stockholm University, S-106 91 Stockholm, Sweden. 3 Department of Botany, Box 355325, University of Washington, Seattle,Washington 98195-5325, U.S.A. Harvard Papers in Botany, Vol. 6,No. 2, 2001, pp. 459–479. © President and Fellows of Har vard College, 2001. Several studies of Dipsacales phylogeny have appeared in recent years, based on both mor- phological and molecular evidence (Donoghue, 1983; Donoghue et al., 1992; Judd et al., 1994; Backlund and Donoghue, 1996; Backlund and Bremer, 1997; Pyck et al., 1999; Pyck and Smets, 2000; Olmstead et al., 2000). These have concluded that Viburnum is related to Sambucus plus Adoxa and that the remainder of the traditional Caprifoliaceae are more closely related to Morinaceae, Valerianaceae, and Dipsacaceae. Furthermore, these studies have supported the monophyly of Caprifolieae, Diervilleae, and Linnaeeae (all of traditional Caprifoliaceae), and the view that Linnaeeae are more c losely related to Morinaceae, Valerianaceae, and Dipsacaceae. However, sup- port for these clades has not been uniformly strong, and the exact placement of several key taxa (Heptacodium, Sinadoxa, Tetradoxa, Triplostegia) remains uncertain. Here we present an expanded analysis of chloroplast DNA rbcL sequences for Dipsacales, and a combined analysis of rbcL with morphological characters. We focus spe- cial attention on phylogenetic relationships within Adoxaceae, based on a combination of nuclear ribosomal internal transcribed spacer (ITS) and rbcL sequences. Of special interest is the placement of the two rare and morphologi- cally bizarre Chinese species, Sinadoxa cory - dalifolia and Tetradoxa omeiensis. The recent rediscovery of these species allows us to pre- sent molecular evidence on their relationships.

PHYLOGENY AND PHYLOGENETIC TAXONOMY OF DIPSACALES…phylodiversity.net/.../MJD_papers/2001/112_MJD_HarvardPapBot01.pdf · phylogeny and phylogenetic taxonomy of dipsacales, with special

Embed Size (px)

Citation preview

PHYLOGENY AND PHYLOGENETIC TAXONOMY OF DIPSACALES, WITH SPECIAL REFERENCE TO SINADOXA

AND TETRADOXA (ADOXACEAE)

MICHAEL J. DONOGHUE,1 TORSTEN ERIKSSON,2 PATRICK A. REEVES,3AND RICHARD G. OLMSTEAD 3

Abstract. To further clarify phylogenetic relationships within Dipsacales,we analyzed new and previously pub-lished rbcL sequences, alone and in combination with morphological data. We also examined relationshipswithin Adoxaceae using rbcL and nuclear ribosomal internal transcribed spacer (ITS) sequences. We concludefrom these analyses that Dipsacales comprise two major lineages:Adoxaceae and Caprifoliaceae (sensu Judd etal.,1994), which both contain elements of traditional Caprifoliaceae. Within Adoxaceae, the following relation-ships are strongly supported: (Viburnum (Sambucus (Sinadoxa (Tetradoxa, Adoxa)))). Combined analyses ofC ap ri foliaceae yield the fo l l ow i n g : ( C ap ri folieae (Diervilleae (Linnaeeae (Morinaceae (Dipsacaceae(Triplostegia, Valerianaceae)))))). On the basis of these results we provide phylogenetic definitions for the namesof several major clades. Within Adoxaceae, Adoxina refers to the clade including Sinadoxa, Tetradoxa, andAdoxa. This lineage is marked by herbaceous habit, reduction in the number of perianth parts,nectaries of mul-ticellular hairs on the perianth,and bifid stamens. The clade including Morinaceae, Valerianaceae, Triplostegia,and Dipsacaceae is here named Valerina. Probable synapomorphies include herbaceousness,presence of an epi-calyx (lost or modified in Valerianaceae), reduced endosperm,and distinctive chemistry, including production ofmonoterpenoids. The clade containing Valerina plus Linnaeeae we name Linnina. This lineage is distinguishedby reduction to four (or fewer) stamens, by abortion of two of the three carpels,and possibly by supernumeraryinflorescences bracts.

Keywords: Adoxaceae, Caprifoliaceae, Dipsacales, ITS, morphological characters, phylogeny, phylogenetictaxonomy, phylogenetic nomenclature, rbcL, Sinadoxa, Tetradoxa.

We are grateful to Jim Smith for rbcL sequencing efforts early in the life of this project, and to Bruce Baldwin for helpwith ITS sequencing. We also thank D. Boufford, B. Tan, K. In (Chengdu, Sichuan), T.-S. Ying (Beijing), and T.-N. Ho(Xining, Qinghai) for organizing expeditions to re-collect Tetradoxa and Sinadoxa, and for their help in the field. The expedition to Qinghai was supported by a National Geographic Society award to D. Boufford. Nancy Pyck and Chuck Bellshared information on their ndhF studies of Dipsacales, and Anders Backlund provided useful comments early on.

1Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut, 06511-8160 U.S.A. E-mail: [email protected]

2Bergius Foundation, Royal Swedish Academy of Sciences, Box 50017, S-104 05 Stockholm, Sweden,and Departmentof Botany, Stockholm University, S-106 91 Stockholm, Sweden.

3Department of Botany, Box 355325, University of Washington, Seattle, Washington 98195-5325, U.S.A.

Harvard Papers in Botany, Vol. 6,No. 2, 2001, pp. 459–479.© President and Fellows of Harvard College, 2001.

Several studies of Dipsacales phylogeny haveappeared in recent years, based on both mor-phological and molecular evidence (Donoghue,1983; Donoghue et al., 1992; Judd et al., 1994;Backlund and Donoghue, 1996; Backlund andBremer, 1997; Pyck et al., 1999; Pyck andSmets, 2000; Olmstead et al., 2000). Thesehave concluded that Viburnum is related toSambucus plus Adoxa and that the remainder ofthe traditional Caprifoliaceae are more closelyre l ated to Mori n a c e a e, Va l e ri a n a c e a e, a n dDipsacaceae. Furthermore, these studies haves u p p o rted the monophy ly of C ap ri fo l i e a e,Diervilleae, and Linnaeeae (all of traditionalCaprifoliaceae), and the view that Linnaeeaea re more cl o s e ly re l ated to Mori n a c e a e,Valerianaceae, and Dipsacaceae. However, sup-

port for these clades has not been uniformlystrong, and the exact placement of several keytaxa (H ep t a c o d i u m, S i n a d ox a, Te t ra d ox a,Triplostegia) remains uncertain.

Here we present an expanded analysis ofch l o roplast DNA r b cL sequences fo rDipsacales, and a combined analysis of rbcLwith morphological characters. We focus spe-cial attention on phy l ogenetic re l at i o n s h i p swithin Adoxaceae, based on a combination ofnuclear ribosomal internal transcribed spacer(ITS) and rbcL sequences. Of special interest isthe placement of the two rare and morphologi-cally bizarre Chinese species, Sinadoxa cory -dalifolia and Tetradoxa omeiensis. The recentrediscovery of these species allows us to pre-sent molecular evidence on their relationships.

On the basis of these findings we contrast ap hy l ogenetic nomencl at u ral system fo rDipsacales with a traditional Linnaean taxo-nomic treatment, and present phylogenetic def-initions for the names of several major clades.

MATERIALS AND METHODS

TaxaOur rbcL analyses included 30 species of tra-

ditional Dipsacales, covering all major lineages(Table 1). We also included Columellia andDesfontainia, which may be closely related to,or possibly even nested within, D i p s a c a l e s(Backlund and Donoghue, 1996; Backlund andBremer, 1997). Previous analyses have shownt h at Dipsacales belong within A s t e ri d a e(Olmstead et al., 1992, 1993; Chase et al.,1993; Backlund and Bremer 1997), where theyare related to Apiales and Asterales within a“euasterid II” clade (APG, 1998). Here we haveincluded published rbcL sequences from fourspecies of Apiales (Coriandrum, Griselinia,H e d e ra, P i t t o s p o ru m) , and four species ofA s t e rales (B a rn a d e s i a, B o o p i s, C a m p a nu l a,M e nya n t h e s). In addition, we added Gentianaand Nicotiana as representatives of the “euas-terid I” clade (APG, 1998), and Cornus ofCornales. On the basis of all previous phyloge-netic analyses, our trees were rooted along thebranch connecting Cornus.

Our ITS analyses included new sequencesf rom Te t ra d ox a and S i n a d ox a ( Table 1), a l o n gwith five sequences from S a m bu c u s ( E ri k s s o nand Donog h u e, 1997) and five from Vi bu rnu m( D o n oghue and Baldw i n , 1993; Baldwin et al.,1995; see also Eriksson and Donog h u e, 1 9 9 7 ) .Sequences of D i e rv i l l a and We i ge l a ( D i e rv i l l e a e ),obtained from Genbank (Kim and Kim, 1999),were included for rooting purposes.

SequencingOf the 43 rbcL sequences included in our

analyses, 27 were obtained from the authors ofprevious studies or from Genbank (Table 1) and16 are reported here for the first time. The newsequences were obtained using standard proto-cols and primers (see, e.g., Olmstead et al.,1992, 1993). ITS sequences for Sinadoxa andTe t ra d ox a we re obtained using pro t o c o l sdescribed elsewhere (Eriksson and Donoghue,1997); however, we used a different 5' primer(ITS-I of Urbatsch et al., 2000), and reads werelong enough that internal primers were notn e e d e d. All sequences rep o rted here we rep ro o f read using sequences of both DNA

strands. ITS sequences were edited using theStaden package (Staden, 1996) under the Linuxoperating system.

AlignmentrbcL nucleotide sequence data were aligned

manually using the editor in PAUP* (Swofford,2001); no indels were recorded. The rbcL datamatrix contained 2% “missing data” for the 43t a x a , re flecting uncertainty in base-calling.Primer positions 1–26 were excluded from theanalyses. Of the remaining 1402 characters,230 were parsimony informative.

ITS nucleotide sequences from 15 taxa werealigned manually using the Se-Al alignmenteditor (Rambaut, 1 9 9 6 ) , resulting in 679aligned positions. Positions 414–482 we reremoved from parsimony analyses owing todifficulties in alignment. Indels were treated as“uncertain” in the matrix, and 21 parsimony-informative indels (positions 47, 48, 53, 92,111–112, 129, 166, 203, 210–212, 216, 221,222, 235, 372, 375, 405, 572, 581, 607–608,6 0 9 , and 631) we re coded as binarypresence/absence characters and added to theend of the matrix. Our ITS matrix contained6.2% uncertain data: 0.4% due to uncertainscorings, 2.8% to implied indels, and 3.0% tofilling in unsequenced portions of the 5.8S genein several species. In total, our ITS matrix con-tained 181 parsimony-informative characters.

Combining DataWe analyzed five datasets. Three of these

focused on Dipsacales: (1) rbcL alone forDipsacales and outgroups, (2) rbcL plus mor-phological data from Judd et al. (1994), and (3)rbcL plus morphological data from Backlundand Donoghue (1996). All data from the mor-phological studies were added without changesin the characters or in taxon scoring; bothdatasets also included chromosome and sec-ondary chemical characters. Two additionaldatasets focused on Adoxaceae: (4) ITS alonefor Adoxaceae and outgroups, and (5) ITS plusrbcL. Here we describe assembly of the com-bined datasets (2, 3, and 5).

The Judd et al. (1994) morphological analy-sis included a subset of the taxa in our rbcLmatrix. From our rbcL dataset we omitted 12outgroup taxa not present in Judd et al., keep-ing only Pittosporum for rooting purposes.Judd et al. included Alseuosmiaceae, but theseare probably not closely related to Dipsacales(Gustafsson et al., 1996) and have been omitted

460 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

here. We also omitted rbcL sequences for thoseDipsacales not rep resented in Ju dd et al.:Heptacodium, Morina, Sinadoxa, Tetradoxa,and Triplostegia. Judd et al. lumped Weigelaand Diervilla together, and also Kolkwitzia andAbelia, because these scored identically fortheir 35 (25 binary, 10 multistate) morphologi-cal characters. In our analysis each of thesepairs was separated into two taxa, which werescored identically for all morphological charac-ters. Several taxa in Judd et al. were repre-sented by two or more rbcL sequences in ourmatrix. Using the “merge taxa” function inMacClade (Maddison and Maddison, 1992),we merged the sequences of our two Sambucusspecies into a single sequence with six poly-m o rphic positions. Similarly, we merge dsequences of Vi bu rnu m ( t h ree species),L o n i c e ra ( t wo species), Dipsacaceae (thre es p e c i e s ) , and Va l e rianaceae (five species),resulting in 28, 55, 40, and 100 polymorphicpositions, respectively. The resulting combinedmatrix contained 16 taxa and 1437 characters(after excluding primer sites 1–26); there were93 parsimony-informative characters.

The Backlund and Donoghue (1996) datasetcontained many more taxa (58) and characters(109). However, owing to differences in taxonsampling, we removed 9 outgroup taxa fromour r b cL dataset (B a rn a d e s i a, B o o p i s,Campanula, Coriandrum, Cornus, Gentiana,H e d e ra, M e nya n t h e s, and N i c o t i a n a) .Likewise, from the morphological dataset weremoved 12 Dipsacales and 16 outgroup taxa.Sambucus, Viburnum, and Lonicera sequenceswere merged as described above. The com-bined dataset contained 30 taxa and 1511 char-acters (after excluding primer sites 1–26); therewere 236 parsimony-informative characters.

The combined analysis of A d oxaceae used thei n t e rsection of all ava i l able ITS and r b cLsequences. It contained 10 taxa and 2012 ch a ra c-t e rs (after ex cluding primer sites 1–26 and 69 ITSsites in a region of ambiguous alignment); therewe re 229 pars i m o ny - i n fo rm at ive ch a ra c t e rs .

All of our data mat ri c e s , and the tre e srep o rted here, a re ava i l able in Tre e BA S E(http://treebase.org).Phylogenetic Analyses

Parsimony analyses were conducted usingPAUP* (Swofford, 2001). For datasets 1 and 3(see above), we carried out heuristic searches(1000 random addition sequence starting trees,TBR branch swapping, saving all most parsi-monious trees); bra n ch-and-bound search e s

were possible for datasets 2, 4, and 5. All char-acters were weighted equally, and indels weret re ated as described ab ove. MacClade andPAUP* were used for the output of trees, char-acter optimizations, and so on.

Measures of SupportDecay indices (Bremer, 1988; Donoghue et

al., 1992) were calculated using AutoDecay(Eriksson, 1999) and PAUP*. The converseconstraint runs in PAUP* were performed usingbranch-and-bound searches when possible; oth-erwise, heuristic searches were conducted with100 random addition sequence starting trees,TBR branch swapping, and saving all most par-simonious trees. Bootstrap analy s e s(Felsenstein, 1985) were carried out using 500rep l i c ations with simple addition sequence,TBR branch swapping, and saving all most par-simonious trees in each replicate.

RESULTS

Pa rs i m o ny analysis of our r b cL mat ri x(analysis 1) resulted in four minimum-lengthtrees of 996 steps (CI = 0.521; CI excludinguninformative = 0.404; RI = 0.593); these dif-fer only in whether S y m p h o ri c a rp o s o rTriosteum is more closely related to Lonicera(Fig. 1). This result supports many previousconclusions. Dipsacales are seen to be mono-phyletic with Columellia and Desfontainia astheir sister group. This is in contrast to severalp revious rep o rts that C o l u m e l l i a a n dDesfontainia may be nested within the group(e.g., Backlund and Donoghue, 1996; Backlundand Bremer, 1997; see discussion). We find lit-tle support for a connection betwe e nA d oxaceae and A raliales (Backlund andBremer, 1997); forcing these taxa together addsa minimum of five steps. The tra d i t i o n a lDipsacaceae and Valerianaceae are both mono-phyletic, show internal relationships largelyconsistent with previous results (but see Caputoand Cozzollino, 1994, on Dipsacaceae), and areweakly united with Triplostegia. In our rbcLanalysis, it is equally parsimonious to placeTriplostegia as the sister group of Valerianaceae(as in previous studies: B a cklund andDonoghue, 1996; Backlund and Bremer, 1997;see discussion) or Dipsacaceae. M o ri n a( M o rinaceae) is sister to the Dipsacaceae-Triplostegia-Valerianaceae clade.

Caprifoliaceae, in the traditional sense, arecl e a rly not monophy l e t i c. Vi bu rnu m a n dSambucus are united with Adoxa and its rela-

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 461

462 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

FI G U R E 1. Strict consensus of four most parsimonious trees resulting from the analysis of rbcL sequence dataalone (analysis 1); length=996 steps; consistency index (CI), ex cluding uninfo rm at ive ch a ra c t e rs=0.40; re t e n t i o ni n d ex (RI)= 0.59. Bootstrap values gre ater than 50% are indicated ab ove the bra n ches and decay indices below.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 463

tives, with Sinadoxa seen to be the sister groupof A d ox a plus Te t ra d ox a. Cap ri fo l i e a e,Diervilleae, and Linnaeeae, of the traditionalCaprifoliaceae, are more closely related to theM o ri n a c e a e - D i p s a c a c e a e -Tr i p l o s t egi a-Valerianaceae clade. Caprifolieae are mono-phyletic, with Leycesteria the sister group of aclade including S y m p h o ri c a rp o s, Tri o s t e u m,and Lonicera. This clade is in turn weaklyunited with H ep t a c o d i u m. Dierv i l l e a e(D i e rv i l l a, We i ge l a) are also monophy l e t i c.However, in contrast with previous analyses,Linnaeeae do not form a clade in this analysis.Linnaea is separated from Abelia, Dipelta, andKolkwitzia, which are in turn paraphyletic inrelation to Diervilleae. Branch lengths in thisportion of the tree are especially short (Fig. 2),i m p lying that other arra n gements may benearly as parsimonious. Trees constrained tounite Abelia, Dipelta, and Kolkwitzia are onlytwo steps longer than our shortest trees; how-ever, linking Linnaea to this group, in a mono-phyletic Linnaeeae, yields trees that are fivesteps longer.

Parsimony analyses of rbcL sequences com-bined with morphological characters yield sev-eral important insights. Combining rbcL withthe Judd et al. (1994) morphological dataset(analysis 2) yielded five trees of 258 steps (Fig.3; CI = 0.767; CI excluding uninformative =0.657; RI = 0.757). These trees are entirelycongruent with those obtained by Judd et al.(1994) using morphology alone, and in almostevery respect they are also congruent with therbcL trees discussed above. However, in con-t rast to results from analyses using r b cLsequences alone, the Linnaeeae form a weaklys u p p o rted cl a d e, wh i ch is sep a rated fro mDiervilleae and instead united directly withDipsacaceae and Valerianaceae.

Combining r b cL with the Backlund andD o n oghue (1996) morp h o l ogical ch a ra c t e rs(analysis 3) yielded 12 trees of 861 steps (Fig.4; CI = 0.569; CI excluding uninformative =0.478; RI = 0.689). Almost all of the clades inthese trees correspond with those found usingm o rp h o l ogical ch a ra c t e rs alone, r b cLsequences alone, and rbcL plus the Judd et al.morphological characters. The main differenceis between this combined analysis and the rbcLsequences analyzed separately. Although sup-port values are not high, here again the com-bined data yield a monophyletic Linnaeeae,which is separated from Diervilleae and unitedwith a clade including Mori n a c e a e,

Valerianaceae, Triplostegia, and Dipsacaceae.The combined analysis also united Triplostegiawith Va l e ri a n a c e a e, instead of withDipsacaceae as in the rbcL analysis.

Our analysis of Adoxaceae based on ITSsequences alone (analysis 4) resulted in twotrees of 347 steps (Fig. 5; CI = 0.784; CIexcluding uninformative = 0.750; RI = 0.864).As in previous analyses (Donog h u e, 1 9 8 3 ;D o n oghue et al., 1992; Ju dd et al., 1 9 9 4 ;B a cklund and Donog h u e, 1 9 9 6 ) , Vi bu rnu mappears to be the sister group of a clade con-taining Sambucus and Adoxa. The significantnew result is strong support for the placementof S i n a d ox a and Te t ra d ox a in re l ation toAdoxa. As in the rbcL analyses just discussed,Sinadoxa is the sister group of a Tetradoxa-Adoxa clade (also see Backlund and Donoghue,1996). As expected, the combined analysis ofITS and rbcL sequences for Adoxaceae (analy-sis 5) yielded a single tree of 395 steps (CI =0.858; CI excluding uninformative = 0.828; RI= 0.881) with remarkably high support valuesfor all clades (Fig. 6).

DISCUSSION

Phylogenetic RelationshipsWith the availability of one or more rbcL

sequences for all of the standardly recognizedge n e ra of the traditional Cap ri fo l i a c e a e(Zabelia, a segregate of the traditional Abelia,is not included here), and a reasonable sampleof Va l e rianaceae and Dipsacaceae, m a nyaspects of the phylogeny of Dipsacales are nowestablished with considerable certainty. Mosti m p o rt a n t ly, it is cl e a r, as highlighted byDonoghue (1983), that the traditional circum-scription of Caprifoliaceae must be abandoned(also see Donoghue et al., 1992; Judd et al.,1994; Backlund and Donoghue, 1996; and laterauthors). Viburnum and Sambucus are moreclosely related to Adoxa and its relatives thanthey are to the other traditional Caprifoliaceae.Caprifolieae, Diervilleae, and Linnaeeae, of thet raditional Cap ri fo l i a c e a e, a re more cl o s e lyre l ated to Mori n a c e a e, Va l e ri a n a c e a e,Triplostegia, and Dipsacaceae than they are toVi bu rnu m and S a m bu c u s. Of these gro u p s ,Linnaeeae appear to be more closely related tothe Mori n a c e a e - Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae clade than to Cap ri folieae orDiervilleae.

For purposes of the remaining discussion wewill adopt the names for the two major clades ofDipsacales suggested by Ju dd et al. (1994).

464 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

FI G U R E 2. Phylogram from the rbcL analysis; branch lengths scaled to the number of inferred state changes.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 465

FI G U R E 3. Strict consensus of five most parsimonious trees resulting from the analysis of rbcL sequence datacombined with the morphological characters from Judd et al. (1994) (analysis 2); length=258 steps; consis-tency index (CI), excluding uninformative characters=0.66; retention index (RI)= 0.76. Bootstrap valuesgreater than 50% are indicated above the branches and decay indices below.

466 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

FI G U R E 4. Strict consensus of 12 most parsimonious trees resulting from the analysis of rbcL sequence datacombined with the morphological characters from Backlund and Donoghue (1996) (analysis 3); length=861steps; consistency index (CI), excluding uninformative characters=0.48; retention index (RI)=0.69. Bootstrapvalues greater than 50% are indicated above the branches and decay indices below.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 467

FI G U R E 5. Phylogram of one of the two most parsimonious trees resulting from the analysis of Adoxaceaebased on ITS sequences (analysis 4); length=347 steps; consistency index (CI), excluding uninformative char-acters=0.75; retention index (RI)=0.86. Bootstrap values greater than 50% are indicated above the branchesand decay indices below. Branch lengths scaled to the number of inferred state changes.

468 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

FI G U R E 6. The single most parsimonious trees resulting from the analysis of Adoxaceae based on rbcL and ITSsequences (analysis 5); length=395 steps; consistency index (CI), excluding uninformative characters=0.83;retention index (RI)=0.88. Bootstrap values greater than 50% are indicated above the branches and decayindices below.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 469

A d oxaceae will re fer to the clade incl u d i n gVi bu rnu m, S a m bu c u s, S i n a d ox a, Te t ra d ox a, a n dA d ox a. Cap ri foliaceae will re fer to the cl a d ei n cluding Cap ri fo l i e a e, D i e rv i l l e a e, a n dLinnaeeae (of traditional Cap ri fo l i a c e a e ) , p l u sM o ri n a c e a e, Va l e ri a n a c e a e, Tri p l o s t egi a, a n dD i p s a c a c e a e. Note especially that Cap ri fo l i a c e a es e n s u Ju dd et al. (1994) includes seve ral taxat ra d i t i o n a l ly tre ated as families (e. g. ,D i p s a c a c e a e, Va l e ri a n a c e a e ) , whose names weretain here (see below under “ P hy l oge n e t i cTa x o n o my”). A d ox a c e a e, C ap ri fo l i a c e a e, a n dthe names of other major clades are given fo rm a lp hy l ogenetic definitions below.

Relationships and evolution in Adoxaceae.Relationships within Adoxaceae now seem

firmly established. Sambucus and Adoxa havepreviously been united on the basis of morpho-logical characters (Donoghue, 1983; Judd eta l . , 1994; Backlund and Donog h u e, 1 9 9 6 ) ,rbcL sequences (e.g., Donoghue et al., 1992;Chase et al., 1993; Backlund and Bre m e r,1 9 9 7 ) , and ITS sequences (Eriksson andD o n og h u e, 1997). The re l ationships ofS i n a d ox a and Te t ra d ox a h ave, h oweve r,received rather little attention, owing in largepart to the rarity of these plants and, therefore,the limited material available for study (but seeLiang, 1997). These taxa were both describedfrom China in 1981—Sinadoxa by Wu et al.(1981) from limestone outcrops near Nangqenin the southeast corner of Qinghai province,and Tetradoxa by Hara (1981; Wu, 1981) fromMt. Omei in Sichuan prov i n c e. Relocat i n gthese plants has proven difficult, however, andfew additional specimens have come to light. In1995 an expedition organized by Dr. In Kaipuof the Chengdu Botanical Institute was suc-cessful in re d i s c ove ring Te t ra d ox a on Mt.Omei. Later that summer, Sinadoxa was recol-lected in southern Qinghai on an expeditionj o i n t ly orga n i zed by the Xining BotanicalInstitute, the Beijing Botanical Institute, andthe Arnold Arboretum of Harvard University.Both species appear to be extremely rare, butsufficient material was obtained to conduct thesequencing reported here (Table 1).

S eve ral authors have argued that A d ox a a n dS i n a d ox a a re dire c t ly re l at e d, with Te t ra d ox am o re distant and retaining the largest number ofa n c e s t ral fe at u res (Wu , 1981; Liang and Zhang,1986; Liang, 1997). These three taxa fo rmed aclade in the morp h o l ogical analysis of Back l u n dand Donoghue (1996), but re l ationships among

them we re unre s o l ve d. All of our moleculara n a ly s e s , based on r b cL alone, ITS alone, a n dthe two combined, s u p p o rt the S i n a d ox a-Te t ra d ox a-A d ox a cl a d e, with bootstrap values of100% in each case, and decay values of 10 fo rr b cL , 34 for ITS, and 42 for the combined dat a .These analyses also support the conclusion thatA d ox a and Te t ra d ox a a re more cl o s e ly re l ated toone another than either is to S i n a d ox a. Bootstrapand decay values are high for the Te t ra d ox a-A d ox a cl a d e : 89% and decay = 2 for r b cL alone,100% and decay = 8 for ITS alone, and 100%and decay = 9 for the combined analysis.

These results bear importantly on our inter-pretation of character evolution in Adoxaceae.The fi rst split, b e t ween Vi bu rnu m and theS a m bu c u s-A d ox a clade (Adoxoideae s e n s uThorne, 1992; Eriksson and Donoghue, 1997)is marked by several morphological changes.Abortion of two of the three carpels and dis-placement of the remaining fertile ovule into as t e rile carpel (Wi l k i n s o n , 1 9 4 8 , 1 9 4 9 ;Fukuoka, 1972) presumably evolved along theViburnum branch (Donoghue, 1983); the result-ing one-seeded drupe is also pro b ably aViburnum synapomorphy. The Adoxoideae aremarked by the evolution of pinnately com-pound (or deeply dissected) leaves, simple ves-sel perforations, extrorse anthers (at least atmaturity; see Erbar, 1994), vestigial archespo-rial tissue (Erbar, 1994), Adoxa-type embryosac deve l o p m e n t , and possibly by seve ra lchemical characters, including production ofalkaloids (Donoghue, 1983; Judd et al., 1994;Backlund and Donoghue, 1996; Eriksson andDonoghue, 1997).

Within Adoxoideae, exactly which morpho-logical changes mark the Sambucus branchremains uncertain, though candidate synapo-morphies include crystal sand in cortical cellsand apical opening of the endocarp (seeEriksson and Donoghue, 1997). In contrast, theS i n a d ox a-Te t ra d ox a-A d ox a clade is we l lmarked by the evolution of the herbaceoushabit, with rhizomes and a basal rosette ofdeeply lobed leaves; reduction of the perianthparts to four or fewer in at least some flowers;nectaries of multicellular hairs positioned onthe perianth parts (see Wagenitz and Laing,1984; Erbar, 1994); bifid stamens (arising froma single primordium) in which the filaments aremore or less deeply split and the monothecousanthers are therefore widely separated; separatestyles with small stigmas; and fruits that arem o re - o r-less dry at mat u rity (possibly dis-

persed by ants). These plants may also have adistinctive chemistry, including loss of valeri-anic acid and the production of coumarins,though this has not been studied in Sinadoxaand Tetradoxa. It is noteworthy that several ofthese features appear to have evolved indepen-dently within Sambucus, including herbaceous-ness, reduction in the number of perianth parts,and a tendency toward splitting of the stamensand styles (Eriksson and Donoghue, 1997).

Within the Sinadoxa-Tetradoxa-Adoxa clade,Sinadoxa is unique by virtue of its spikelikeinflorescence structure (with scattered clustersof three to five flowers), extreme variation andreduction in the number of perianth parts, andespecially by the reduction to just a singlecarpel and the production of fruits with calyx-d e rived carnose sacs (wh i ch may fa c i l i t at ewater dispersal). The Tetradoxa-Adoxa clade ischaracterized by smaller plants with leaves thatare trifoliate or ternately lobed. In Tetradoxa,the perianth parts are acuminate at the apex (asthey also are in several Sambucus species,including S. ebulus; Eriksson and Donoghue,1997); fusion of filament-halves at the base(i.e., separation of filaments above the middle)may also be an apomorphy of this species. Theelongate inflorescence of Tetradoxa, with flow-ers on distinct pedicels, may represent theancestral condition from which the headlikeinflorescence of Adoxa was derived by “com-pression.” Although most Tetradoxa flowers are4-merous, they show considerable variation inmerousness along the inflorescence axis (Liangand Zhang, 1986); in Adoxa, the four lateralflowers in the cubelike head are typically five-merous (usually with only three sepals devel-oped), and the terminal flower is four-merous(see Erbar, 1994).

Liang (1997) showed that Te t ra d ox a a n dSinadoxa have simpler floral vasculature thaneither Adoxa moschatellina or A. orientalis(Nepomnyashchaya, 1984) and suggested thatTetradoxa, in particular, had retained the ances-tral characteristics (see also Liang and Zhang,1986). Our results suggest that elongate inflo-rescences may have been ancestral, but it isp o s s i ble that the T- s h aped stamens ofTetradoxa were derived from an ancestor withthe completely bifid condition, as seen in theother species. Inclusion of S i n a d ox a a n dTetradoxa in developmental studies will benecessary to determine the generality of manyother unusual characteristics seen in Adoxa(Erbar, 1994), such as “lobelioid” orientation of

the five-merous flowers (also see Donoghue eta l . , 1 9 9 8 ) , loss of early corolla ring pri-mordium, and reduced vestigial archesporium.

Relationships and evolution in Caprifoliaceae.Comparison of our separate and combined

analyses shows that, although these generallyyield similar results, morphological charactersare important in resolving a number of rela-tionships. Perhaps most importantly, morpho-l ogical data tend to unite the ge n e ra ofLinnaeeae, greatly increase support for theM o ri n a c e a e - Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae clade, and link these two cladesd i re c t ly toge t h e r. Th ey also seem to linkTriplostegia with Valerianaceae, as opposed toDipsacaceae.

Linnaeeae we re united by supernu m e ra ryi n fl o rescence bracts (see Fukuoka, 1 9 6 9 ;Web e rl i n g, 1989) in Ju dd et al. (1994).However, it is perhaps more likely that this fea-ture is characteristic of Linnaeeae plus theM o ri n a c e a e - Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae clade. Under this view, supernu-merary bracts were modified and fused to formthe characteristic epicalyx seen in Morinaceaeand Dipsacaceae (Hofmann and Göttmann,1990; Manchester and Donoghue, 1995; Roelsand Smets, 1996; Backlund and Donoghue,1996). This scenario would, however, requirethe addition of a second ep i c a lyx inTri p l o s t egi a and the loss of ep i c a lyx inValerianaceae (or perhaps its modification toform fruit wings, as in Patrinia). It is possiblethat Linnaeeae are distinguished by supervoluteleaf vernation and by contraction of the chloro-plast DNA inverted repeat, but data are still toolimited to be certain (Backlund and Donoghue,1996). The best support for the monophyly ofLinnaeeae comes from analyses of ndhF andother chloroplast genes. Studies by Pyck et al.(1999), Pyck and Smets (2000), and Bell et al.(2001) all support Linnaeeae as a clade.

M o rp h o l ogical data provide more conv i n c i n gs u p p o rt for the Mori n a c e a e - Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae cl a d e, wh i ch is onlywe a k ly supported in r b cL analyses. A nu m b e rof ch a ra c t e rs re flect a shift to herbaceousness inthis gro u p , i n cluding the presence of a tap ro o tand persistent basal leaves (in addition tocauline leave s ) , the absence of bud scales, and at e n d e n cy towa rd simple perfo ration plates in thevessels (especially in Va l e rianaceae). Po s s i bl erep ro d u c t ive synap o m o rphies include the pre s-ence of an ep i c a lyx (though, as noted ab ove, t h i s

470 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

would re q u i re loss in Va l e ri a n a c e a e ) , re d u c t i o nof the calyx lobes or modifi c ation into pers i s t e n tb ristles or pappus (more “ n o rm a l ” lobes are seen i nM o ri n a c e a e a n din N a rd o s t a chy so f Va l e ri a n a c e ae) ,and reduction in endosperm , c u l m i n ating in com-p l e t e loss in Va l e ri a n a c e a e. Fi n a l ly, the Mori n a c e a e-Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae cl a d eis ch a ra c t e ri zed by distinctive ch e m i s t ry, i n cl u d i n gthe presence of alkaloids, c athecolic tannins,and monoterpenoids.

The link between the Linnaeeae and theM o ri n a c e a e - Va l e ri a n a c e a e -Tri p l o s t egi a-Dipsacaceae clade is also supported by severalrather cl e a r-cut morp h o l ogical fe at u re s , a shighlighted by Judd et al. (1994). Stamen num-ber is reduced from five to four, which thenmay be didynamous in arrangement. Duringovary development two of the three carpelsabort, leaving a single-seeded carpel occupyinghalf of the ovary. Wilkinson (1949), in particu-lar, stressed the great similarity of this condi-tion in Linnaeeae and Valerianaceae, alongwith the similar median position and compoundvasculature of the ovule (also see Fukuoka,1972). Similar carpel ab o rtion occurs inMorinaceae and in Triplostegia. Under the viewthat this characterizes the entire clade, the lossof a clear sign of ab o rted carpels inDipsacaceae is interpreted as a further modifi-cation. Finally, the ovary in all of these plantsmatures into an achene.

These results have especially import a n timplications for our understanding of ovary andfruit evolution. Trees based on rbcL sequencesalone indicate that Linnaeeae are not mono-phyletic and that Diervilleae were, in effect,derived within Linnaeeae. Such trees imply thatthe bi-carpellate, many-seeded, elongate cap-sules ch a ra c t e ristic of the Diervilleae we red e rived from the tri - c a rp e l l at e, o n e - s e e d e d,achene fruits seen in Linnaeeae, Morinaceae,Tri p l o s t egi a, Va l e ri a n a c e a e, a n d, p ro b ablyancestrally, in Dipsacaceae. As we have noted,this result is not upheld in combined analyseswith morphological characters; Linnaeeae areinstead linked with the Mori n a c e a e -Valerianaceae-Triplostegia-Dipsacaceae clade.Support for this result is not especially strong,but it cert a i n ly yields a simpler scenario for fru i tevolution. Specifi c a l ly, our combined trees implythat abortion of two carpels and achene fruitsevo l ved once, at the base of the Linnaeeae-M o ri-na c e a e - Va l e rianaceae Tri p l o s t egi a- D i p s a c a c e a eclade, and did not secondarily give rise to bi-carpellate capsules.

Triplostegia has sometimes been linked withValerianaceae and sometimes with Dipsacaceae(see Backlund and Bremer, 1998). In an earlierrbcL study (Backlund and Bremer, 1998) theposition of Triplostegia was poorly resolved;s u c c e s s ive - ap p rox i m ations weighting we a k lylinked it with Valerianaceae. In our rbcL analy-sis it is we a k ly linked with Dipsacaceaeinstead. Taking morphological characters intoaccount (Backlund and Donog h u e, 1 9 9 6 ) ,Triplostegia is linked with both Valerianaceaeand Dipsacaceae on the basis of simple perfo-ration plates (or nearly so), further reduction ofthe calyx lobes, several pollen characters, andthe presence of ch l o ro p hyllous embryo s(Yakovlev and Zhukova, 1980). Within thisclade, Triplostegia is united with Valerianaceaeon the basis of pollen morphological charac-t e rs , i n cluding ap e rt u res with a distinctive“ h a l o ” ( B a cklund and Nilsson, 1 9 9 7 ) ;endosperm reduction in Triplostegia (Peng etal., 1995) or complete loss in Valerianaceae;and chemical characters, especially the pres-ence of va l ep o t ri ate iridoid compounds(Backlund and Moritz, 1996).

Dipsacales monophyly.Several phylogenetic problems are not con-

v i n c i n g ly re s o l ved by our analy s e s , m o s tnotably the question of the monophyly of tradi-tional Dipsacales (Donoghue, 1983), includingjust Adoxaceae and Caprifoliaceae sensu Juddet al. (1994). In most previous molecular stud -ies, this clade has been recovered (Chase et al.,1993; Olmstead et al., 1992, 1993; Rice et al.,1 9 9 7 ) , though with only weak support .Donoghue et al. (1992) and Downie and Palmer(1992) found alternative arrangements, some-times including Araliaceae, but taxon samplingwas limited in these studies. The main alterna-tive was suggested by rbcL analyses conductedby Bremer et al. (1994), which implied thatDesfontainia may be closely related to, or evennested within, Dipsacales. Outgroups were toopoorly sampled by Judd et al. (1994) to providea critical test using morp h o l ogical dat a .However, the Backlund and Donoghue (1996)morphological analysis, as well as their com-bined analysis with rbcL, showed Columelliaand D e s fo n t a i n i a to be nested withinDipsacales as the sister group(s) of theCaprifoliaceae clade.

In our r b cL analysis the monophy ly ofDipsacales was supported, but only weakly(bootstrap <50%; decay = 1); it requires just a

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 471

single ex t ra step to move the C o l u m e l l i a-Desfontainia clade within the group. With theaddition of the Backlund and Donoghue (1996)m o rp h o l ogical ch a ra c t e rs , t rees of both types areequally parsimonious, that is, with Columelliaand D e s fo n t a i n i a as the sister group of Dipsacalesand with these taxa nested within Dipsacales asthe sister group of Caprifoliaceae.

The original Backlund and Donoghue (1996)analysis, which included a broader sample ofoutgroups, appeared to show much better sup-p o rt for the inclusion of C o l u m e l l i a a n dDesfontainia within Dipsacales. However, oncloser inspection, this is not the case. Both theclade containing the traditional Dipsacales plusColumellia and Desfontainia, and the branchlinking C o l u m e l l i a and D e s fo n t a i n i a w i t hCaprifoliaceae, were very weakly supported(bootstraps <50%; decay = 1). Furthermore,reexamination of the morphological charactersoptimized as changing along these branches(referred to below by character numbers used inBacklund and Donoghue, 1996) shows thatmost of these do not provide unequivocal sup-p o rt. The bra n ch including tra d i t i o n a lDipsacales plus Columellia and Desfontainia ism a rked by 12 morp h o l ogical ch a n ges inBacklund and Donoghue (1996). Two of theseare unknown in Columellia and Desfontainia—iridoids (character 87) and saponins (95)—andfor 7 others the presumed derived state is notp resent in C o l u m e l l i a and/or D e s fo n t a i n i a:l e aves are eve rgreen in C o l u m e l l i a a n dDesfontainia (4); leaf margins in Dipsacalesare very rarely spiny, as in Columellia andDesfontainia (11); leaf vernation is condupli-c ate in C o l u m e l l i a and D e s fo n t a i n i a ( 1 5 ) ;Desfontainia has perforation plates with manycross bars (24); rotate corollas are not found inC o l u m e l l i a and D e s fo n t a i n i a, or inCaprifoliaceae (42); stamens are clearly fusedto the corolla in Columellia and Desfontainia,as in Caprifoliaceae (50); chromosome num-bers are X = 7 in both, not X = 8 as inDipsacales (103). This leaves tenuinucellateovules (65), opposite leaves (5), and petalsfused into a tube (40), all of which are wide-spread among asterids, such that the location ofcharacter changes are heavily dependent onwhich outgroups are included in an analysis.

Th eb ra n ch linking C o l u m e l l i a and D e s fo n t a i n i awith Caprifoliaceae is marked by changes in 14characters in Backlund and Donoghue (1996).However, 4 of these characters are unknown inColumellia and Desfontainia (93, ellagic acid;

1 0 4 , ch romosome size; 106, s t ru c t u ra lre a rra n gement of the ch l o roplast inve rt e drepeat; 108, size of the inverted repeat), andColumellia or Desfontainia lack the presumedderived state of 5 others (27, superficial phel-logen is not present in Desfontainia; 37, threevascular traces are not present in the calyxlobes of Columellia; 46, corolla vasculature isnot laterally connected in Columellia; 59, stig-mas are bilobed, not capitate, in Columellia,and are polymorphic in Desfontainia; 69, thet apetum is glandular, not amoeb o i d, i nDesfontainia). Of the remaining 5 characters,vascular cylinder (21) was scored incorrectlyfor Viburnum and Sambucus (they do havecylinders); well-developed corolla tubes (42)and stamens attached within (50) appear to bedirectly correlated and are widespread amongasterids; corolla zygomorphy (43) is polymor-phic in most genera of Caprifoliaceae s.s.; andthe production of cathecolic tannins (99) ispoorly known and highly variable among pos-sible outgroups and must also be homoplasticwithin Dipsacales.

We conclude from this analysis that molecu-lar and morphological evidence is unconvinc-ing that Columellia and Desfontainia are unitedwith, or nested within, Dipsacales. These taxamay be linked with Dipsacales on the basis ofopposite leaves, fused petals, and tenuinucel-late ovules, and they may specifically be unitedwith Caprifoliaceae on the basis of elongatecorolla tubes bearing the stamens. However,these ch a ra c t e rs each show high levels ofhomoplasy among asterids, and there are obvi-ous conflicting characters. Unlike traditionalDipsacales, Columellia and Desfontainia havespiny, evergreen leaves, bi-carpellate ovariesthat are half inferior in Columellia and fullysuperior in Desfontainia, solanad embryogeny,and base chromosome numbers of X = 7.Placement within Dipsacales also seems oddfrom a biogeographic standpoint; traditionalDipsacales are basically Northern Hemispherein distribution,with greatest diversity in easternAsia, whereas Columellia and Desfontainia arecentered in South America. Finally, traditionalDipsacales do share a number of possible apo-morphies, which may not have been correctlyrepresented in prior analyses, for example, thewidespread occurrence of aborted ovules orvestigial archesporial tissue in the upper por-tion of the ovary (see Erbar, 1994). Dependingon details of broader relationships, cellularendosperm development, asterad and related

472 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

forms of embryogeny, and 3- to 5-carpellategynoecia may also be synapomorphies of tradi-tional Dipsacales.

Phylogenetic TaxonomyIt is possible to reflect our current under-

standing of Dipsacales phylogeny using tradi-tional Linnaean nomenclature; that is, usingstandard ranks such as family, tribe, and so on.This ap p ro a ch has been taken by theAngiosperm Phylogeny Group (APG, 1998),which recognized six families within an OrderDipsacales that excludes Adoxaceae (whichAPG left unassigned within their “euasterid II”clade): (1) Caprifoliaceae (restricted to the for-mer Caprifolieae), (2) Diervillaceae (for theformer Diervilleae; Backlund and Pyck, 1998),(3) Linnaeaceae (for the former Linnaeeae;Backlund and Pyck, 1998), (4) Morinaceae, (5)Dipsacaceae, and (6) Valerianaceae (presum-ably including Tri p l o s t egi a; Backlund andBremer, 1998). Note that in this system oldnames are applied to clades that already havenames (e.g., Caprifoliaceae for Caprifolieae;Dipsacales for Caprifoliaceae sensu Judd et al.,1994), new names are introduced for cladesthat already have names that are in wide use(Diervillaceae for Diervilleae, Linnaeaceae forLinnaeeae), and, ironically, no new names areintroduced to reflect widely accepted knowl-edge of phylogenetic relationships among thesegroups (e.g., the link between Morinaceae,Valerianaceae, and Dipsacaceae, or betweenthis clade and Linnaeeae).

P hy l ogenetic nomencl at u re (s e n s u de Queiro zand Gauthier, 1992, 1994; see also Hibbett andDonoghue, 1998), without ranks, deals withthis problem more efficiently (Fig. 7). Thenames Cap ri fo l i e a e, D i e rv i l l e a e, L i n n a e e a e,D i p s a c a c e a e, and Va l e rianaceae are simplyretained for their respective clades; renamingthese can only cause confusion. Following Juddet al. (1994), the name Cap ri foliaceae isapplied to the clade including all of these taxa.Likewise, the name Adoxaceae is used for theclade including Vi bu rnu m, S a m bu c u s, a n dAdoxa, and Adoxoideae is used for the cladeincluding Sambucus and Adoxa (Donoghue,1983; Thorne, 1992; Eriksson and Donoghue1997). On the basis of arguments presentedabove, the name Dipsacales is retained for the

clade that includes A d oxaceae andCaprifoliaceae. Although we consider it doubt-ful on present evidence, we recognize the pos-sibility that the least inclusive clade includingA d oxaceae and Cap ri foliaceae might alsoinclude Columellia and Desfontainia.

With phy l ogenetic nomencl at u re, n ew names aregiven only to clades that have not prev i o u s ly beenn a m e d. Within A d ox a c e a e, we here propose thename A d oxina for the clade including S i n a d ox a,Te t ra d ox a, and A d ox a. Within Cap ri foliaceae wep ropose the name Va l e rina for the clade incl u d-ing Mori n a c e a e, D i p s a c a c e a e, Tri p l o s t egi a, a n dVa l e ri a n a c e a e, and the name Linnina is ap p l i e dto the clade including Va l e rina plus Linnaeeae.Fo rmal phy l ogenetic definitions of these cl a d e sa re presented in Table 2.

CONCLUSIONS

P revious phy l ogenetic conclusions arel a rge ly confi rmed by our analy s e s .Caprifoliaceae in the traditional sense are notmonophyletic, and there are two major lineageswithin Dipsacales: (1) Adoxaceae, includingVi bu rnu m, S a m bu c u s, S i n a d ox a, Te t ra d ox a,and Adoxa; and (2) Caprifoliaceae, includingC ap ri fo l i e a e, D i e rv i l l e a e, L i n n a e e a e,Morinaceae, Valerianaceae, Triplostegia, andDipsacaceae. Within Adoxaceae we concludethat Sinadoxa and Tetradoxa are linked withAdoxa, in Adoxina. Within Caprifoliaceae ourcombined analyses support Valerina, includingM o ri n a c e a e, D i p s a c a c e a e, Tri p l o s t egi a, a n dValerianaceae, and Linnina, including Valerinaplus Linnaeeae. Each of these clades is markedby morphological synapomorphies.

Additional data are needed to test the mono-phyly of Linnaeeae, as well as to clarify there l ationships of Diervilleae (whether withC ap ri folieae or with Linnina). A l t h o u g hH ep t a c o d i u m ap p e a rs to be re l ated toC ap ri folieae (rather than Linnaeeae), a n dTri p l o s t egi a m ay be we a k ly united withValerianaceae (rather than Dipsacaceae), welook forward to additional evidence on theseissues. Relationships within Diervilleae haverecently been analyzed (Kim and Kim, 1999),but re l ationships within Cap ri folieae andLinnaeeae remain poorly resolved. Finally, thequestion of the monophy ly of Dipsacales,whether including or excluding Columellia andDesfontainia, needs further attention.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 473

474 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

FI G U R E 7. Summary Dipsacales phylogeny showing the proposed application of names to clades. Phylogeneticdefinitions are provided in Table 2 for those clades marked with a dot.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 475

SPECIES REFERENCE OR VOUCHER GENBANK NO.

Abelia x grandiflora Rehder Cult. Univ. Arizona, Tucson, 1993; AJ420875Donoghue, voucher lacking

Adoxa moschatellina L. rbcL: Donoghue et al., 1992 L01884ITS: Eriksson and Donoghue, 1997 U88194

Barnadesia caryophylla Olmstead et al., 1992 L01887(Vell.) S. F. Blake

Boopis anthemoides Juss. Olmstead et al., 1993 L13860

Brunia albiflora Phillips Backlund and Bremer, 1997 Y10674

Campanula ramulosa Wall. Olmstead et al., 1992 L13861

Centranthus ruber (L.) DC. Cult. Univ. Arizona, Tucson, 1993; AJ420879Donoghue, voucher lacking

Columellia oblonga Ruiz & Pav. Backlund and Bremer, 1997 Y10675

Coriandrum sativum L. Olmstead et al., 1992 L11676

Cornus kousa Hance Olmstead et al., 1993 L14395

Desfontainia spinosa Ruiz & Pav. Bremer et al., 1994 Z29670

Diervilla sessilifolia Buckley rbcL: Bremer et al., 1994 Z29672ITS: Kim and Kim, 1999 AF078703

Dipelta floribunda Maxim. Cult. Arnold Arboretum, 14514-B; AJ420876Kelly and Buckland 32

Dipsacus sativus (L.) Honck. Olmstead et al., 1992 L13864

Gentiana procera Holm Olmstead et al., 1993 L14398

Griselinia lucida G. Forst. Xiang et al., 1993 L11225

Hedera helix L. Olmstead et al., 1992 L01924

Heptacodium miconioides Rehder Cult. Arnold Arboretum 1549-80-D; AJ420873Koller s.n., 12 Oct. 1984 (A)

Knautia intermedia Pernh. & Wettst. Backlund and Bremer, 1997 Y10698

Kolkwitzia amabilis Graebn. Cult. Arnold Arboretum 18090; AJ420877Elsik and Siegel 1558 (A)

Leycesteria formosa Wall. Cult. Kew Botanic Gardens, UK, 1990; AJ420872Donoghue, voucher lacking

Linnaea borealis L. Door County, WI, 1990; AJ420878Donoghue, voucher lacking

Lonicera orientalis Lam. Gustafsson et al., 1996 X87389

Lonicera prolifera (G.Kirchn.) Rehder Cult. Arnold Arboretum, 870-74-A; AJ420870voucher lacking

Menyanthes trifoliata L. Olmstead et al., 1993 L14006

Morina coulteriana Royle Backlund and Bremer, 1997 Y10706

Nardostachys jatamansii (D. Don) DC. Backlund and Bremer, 1997 Y10705

TA B L E 1. Vo u cher info rm ation and GenBank accession nu m b e rs for species used in the r b cL and ITS analy s e s .

476 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

SPECIES REFERENCE OR VOUCHER GENBANK NO.

Nicotiana tabacum L. Lin et al., 1986

Patrinia rupestris (Pall.) Dufr. Backlund and Bremer, 1997 Y10704

Phyllactis bracteata Wedd. Backlund and Bremer, 1997 Y10703

Pittosporum japonicum Morgan and Soltis, 1993 L11202Hort. ex C. Presl.

Pterocephalus lasiospermus Backlund and Bremer, 1997 Y10702Link ex Buch

Sambucus australasica (Lindl.) Fritsch ITS: Eriksson and Donoghue, 1997 U41381

Sambucus australis Cham. & Schltdl. ITS: Eriksson and Donoghue, 1997 U88196

Sambucus caerulea Rafinesque rbcL: vic. Tucson,Arizona, 1993; AJ420867Donoghue, voucher lacking

ITS: Eriksson and Donoghue, 1997 U88197

Sambucus ebulus L. ITS: Eriksson and Donoghue, 1997 U88200

Sambucus racemosa L. rbcL: Donoghue et al., 1992 L14066ITS: Eriksson and Donoghue, 1997 U88207

Sinadoxa corydalifolia C. Y. Wu, rbcL: Boufford et al. 26555 (A) AJ420866Z. L. Wu & R. F. Huang ITS: Boufford et al. 26555 (A) AJ419711

Symphoricarpos albus (L.) S. F. Blake Olmstead et al., 1992 L11682

Tetradoxa omeiensis rbcL: Donoghue et al. 4000 (A) AJ420865(H. Hara) C. Y. Wu ITS: Donoghue et al. 4000 (A) AJ419710

Triosteum perfoliatum L. vic. Madison, WI, 1990; AJ420870Donoghue, voucher lacking

Triplostegia glandulifera Wall. ex DC. Backlund and Bremer, 1997 Y10700

Valeriana officinalis L. Olmstead et al., 1992 L13934

Viburnum acerifolium L. rbcL: Olmstead et al., 1992 L01959ITS: Cult. Arnold Arboretum 1505-67; AJ420923Elsik 2102 (A) AJ420924

Viburnum dentatum L. ITS: Eriksson and Donoghue, 1997 U88552U88553

Viburnum lentago L. ITS: Eriksson and Donoghue, 1997 U88554U88555

Viburnum plicatum Thunb. rbcL: Cult. Arnold Arboretum 1050-67-A; AJ420868Dwyer et al., 4354 (A) ITS: Cult. Arnold Arboretum 1050-67-A; AJ420925Dwyer et al., 4354 (A) AJ420926

Viburnum sieboldii Miq. rbcL: Cult. Arnold Arboretum 616-6-B; AJ420869Elsik et al. 2640 (A) U88556ITS: Eriksson and Donoghue, 1997 U88557

Weigela hortensis (Sieb. & Zuck.) rbcL: Cult. Arnold Arboretum 1897-77-A; AJ420874C. A. Mey. Kelly and Buckland 28 (A)

ITS: Kim and Kim, 1999 AF078713

TABLE 1. (CONT.)

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 477

CLADE DEFINITION1 NOTES

Dipsacales Dipsacus fullonum L. Sensu Donoghue (1983)Linnaea borealis L. and Judd et al. (1994)Lonicera caprifolium L.Viburnum lantana L.

Adoxaceae Viburnum lantana L. Sensu Donoghue (1983) and ThorneSambucus nigra L. (1983); defined phylogenetically byAdoxa moschatellina L. Judd et al. (1994)

Adoxoideae Sambucus nigra L. Sensu Thorne (1983)Adoxa moschatellina L.

Adoxina Adoxa moschatellina L. New clade nameTetradoxa omeiensis (H. Hara) C. Y. WuS i n a d oxa cory d a l i fo l i a C .Y. Wu ,Z. L. Wu & R. F. Huang

Caprifoliaceae Lonicera caprifolium L Defined phylogenetically by Judd et al.Diervilla lonicera Miller (1994); Dipsacales sensu APG (1998)Linnaea borealis L.Valeriana pyrenaica L.Dipsacus fullonum L.

Caprifolieae Lonicera caprifolium L. Caprifoliaceae sensu Backlund and PyckSymphoricarpos orbiculata Moench (1998) and APG (1998)Leycesteria formosa Wall.

Diervilleae Diervilla lonicera Miller Diervilliaceae sensu Backlund and Pyckand Weigela japonica Thunb. (1998) and APG (1998)

Linnina Abelia chinensis R. Brown New clade nameLinnaea borealis L.Morina persica L.Valeriana pyrenaica L.Dipsacus fullonum L.

Linnaeeae Abelia chinensis R. Brown Linnaeaceae sensu Backlund and PyckLinnaea borealis L. (1998) and APG (1998)Kolkwitzia amabilis Graebnerand Dipelta floribunda Maxim.

Valerina Morina persica L. New clade nameValeriana pyrenaica L.Dipsacus fullonum L.

1 All of the definitions are "node-based" (de Queiroz and Gauthier, 1992, 1994; see the PhyloCode athttp://www.ohiou.edu/phylocode/); each name is defined as the least-inclusive clade containing the taxa listed.

TABLE 2. Phylogenetic definitions for clade names in Fig. 7.

ANGIOSPERM PHYLOGENY GROUP (APG). 1998. Anordinal classification for the families of floweringplants. Ann. Missouri Bot. Gard. 85: 531–553.

BACKLUND, A., AND B. BREMER. 1997. Phylogeny ofthe Asteridae s. str. based on rbcL sequences, withparticular reference to the Dipsacales. Pl. Syst.Evol. 207: 225–254.

BACKLUND, A., AND K. BREMER. 1998. To be or notto be—principles of classification and monotypicplant families. Taxon 47: 391–400.

BACKLUND, A., AND M. J. DONOGHUE. 1996. Mor-phology and phylogeny of the order Dipsacales.Pages 1–55 in A. BACKLUND, ED., Phylogeny of theDipsacales. Part IV. Department of SystematicBotany, Uppsala University, Uppsala.

BACKLUND, A., AND T. M ORITZ. 1996. Phylogeneticimplications of an expanded valepotriate distribu-tion in the Va l e ri a n a c e a e. Pages 1–27 in A .BACKLUND, ED., Phylogeny of the Dipsacales. PartIII. Department of Systematic Botany, UppsalaUniversity, Uppsala, Sweden.

BACKLUND, A., AND S. NILSSON. 1997. Pollen mor-p h o l ogy and the systematic position ofTriplostegia (Dipsacales). Taxon 46: 21–31.

BACKLUND, A., AND N. PYCK. 1998. Diervilliaceaeand Linnaeaceae, two new families of caprifo-lioids. Taxon 47: 657–661.

BALDWIN, B. G., M. J. SANDERSON, J. M. PORTER, M.F. WOJCIECHOWSKI, C. S. CAMPBELL, AND M. J.DONOGHUE. 1995. The ITS region of nuclear ribo-somal DNA: A valuable source of evidence onangiosperm phylogeny. Annals of the MissouriBotanical Garden 82: 247–277.

BELL, C. D., E J. EDWARDS, S.-T. KIM, AND M. J.DONOGHUE. 2001. Dipsacales phylogeny based onchloroplast DNA sequences. Harvard Pap. Bot. 6:481–499.

BENKO-ISEPPON, A. M., AND W. MORAWETZ. 2000.Viburnales: Cytological features and a new cir-cumscription. Taxon 49: 5–16.

BREMER, K. 1988. The limits of amino acid sequencedata in angiosperm phylogenetic reconstruction.Evolution 42: 795–803.

BREMER, B. R., G. OLMSTEAD, L. STRUWE, AND J. A.SWEERE. 1994. rbcL sequences support exclusionof R e t z i a, D e s fo n t a i n i a, and N i c o d e m i a( B u ddlejaceae) from Gentianales. Plant Systemat i c sand Evolution 190: 213–230.

CAPUTO, G., AND S. COZZOLLINO. 1994. A cladistica n a lysis of Dipsacaceae (Dipsacales). PlantSystematics and Evolution 189: 41–61.

CH A S E, M. W. , E TA L. (41 others). 1993. Phy l oge n e t i c sof seed plants: An analysis of nu cl e o t i d esequences from the plastid gene rbcL. Annals ofthe Missouri Botanical Garden 80: 528–580.

DE QUEIROZ, K., AND J. GAUTHIER. 1992. Phyloge-netic taxonomy. Annual Review of Ecology andSystematics 23: 449–480.

— — —. 1994. Toward a phylogenetic system of bio-l ogical nomencl at u re. Trends in Ecology andEvolution 9: 27–31.

DEVOS, F. 1951. The stem anatomy of some speciesof the Caprifoliaceae with reference to phylogenyand identification of the species. Ph.D. disserta-tion, Cornell University, Ithaca, N.Y.

DO N O G H U E, M. J. 1983. The phy l ogenetic re l at i o n-ships of Vi bu rnu m. Pages 143–166 in N. I. PL AT N I C K

A N D V. A. FU N K, E D S. , A dvances in Cladistics. Vol. 2.Columbia Unive rsity Pre s s , N ew Yo rk .

DO N O G H U E, M. J. , A N D B. G. BA L DW I N. 1 9 9 3 .Phylogenetic analysis of Viburnum based on ribo-somal DNA sequences from the internal tran-scribed spacer regions. Amer. Jour. Bot. 80: 146.

DONOGHUE, M. J.,R. H. REE, AND D. A. BAUM. 1998.Phylogeny and the evolution of flower symmetryin the Asteridae. Trends Plant Sci. 3: 311–317.

DONOGHUE, M. J., R. G. OLMSTEAD, J. F. SMITH, AND

J. D. PALMER. 1992. Phylogenetic relationships ofDipsacales based on rbcL sequences. Annals ofthe Missouri Botanical Garden 79: 333–345.

DOWNIE, S. R., AND J. D. PALMER. 1992. Restrictionsite mapping of the chloroplast DNA invertedrepeat: A molecular phylogeny of the Asteridae.Annals of the Missouri Botanical Garden 79:266–283.

ERBAR, C. 1994. Contributions to the affinities ofAdoxa from the viewpoint of floral development.Bot. Jahrb. Syst. 116: 259–282.

ERIKSSON, T. 1999. AutoDecay. Version 4.0.2 (com-puter program). Distributed by the author. BergiusFoundation, Royal Swedish Academy of Sciences,Stockholm.

ERIKSSON, T., AND M. J. DONOGHUE. 1997. Phyloge-netic re l ationships of S a m bu c u s and A d ox a(Adoxoideae, Adoxaceae) based on nuclear ribo-somal ITS sequences and preliminary morpholog-ical data. Syst. Bot. 22: 555–573.

FELSENSTEIN, J. 1985. Confidence limits on phyloge-nies: An approach using the bootstrap. Evolution39: 783–791.

FU K U O K A, N. 1969. Infl o rescence of Linnaeae (Cap ri-foliaceae). Acta Phytotaxonomica Geobotanica23: 153–162.

— — —. 1972. Taxonomic study of the Cap ri-foliaceae. Memoirs of the Faculty of Science,Kyoto University, Biology Series 6: 15–58.

———. 1974. Floral morp h o l ogy of A d ox am o s ch at e l l i n a. Acta Phytotaxonomica Geobotanica26:65–76.

GUSTAFSON, M . G. H . , A. BAC K L U N D, A N D B. BR E M E R.1996. Phylogeny of the Asterales sensu lato basedon rbcL sequences with particular reference toGoodeniaceae. Plant Syst. Evol. 199: 217–242.

HARA, H. 1981. A new species of the genus Adoxafrom Mt. Omei of China. Journal of JapaneseBotany 56: 271–274.

— — —. 1983. A Revision of the Caprifoliaceae ofJapan with Reference to Allied Plants in OtherDistricts and the Adoxaceae. Academia ScientificBooks, Tokyo.

HI B B E T T, D. S. , A N D M. J. DO N O G H U E. 1 9 9 8 .Integrating phylogenetic analysis and classifica-tion in fungi. Mycologia 90: 347–356.

478 HARVARD PAPERS IN BOTANY Vol. 6, No. 2

LITERATURE CITED

HOFMANN, U., AND J. GÖTTMANN. 1990. Morina L.and Triplostegia Wall ex DC. im Vergleich mitVa l e rianaceae und Dipsacaceae. Botanisch eJahrbücher für Systematik Pflanzengeschichte undPflanzengeographie 111: 499–553.

JUDD, W. S., R. W. SANDERS, AND M. J. DONOGHUE.1994. Angiosperm family pairs: Preliminary phy-logenetic analyses. Harvard Pap. Bot. 5: 1–51.

KIM, Y.-D., AND S.-H. KIM. 1999. Phylogeny ofWeigela and Diervilla (Caprifoliaceae) based onnuclear rDNA ITS sequences: Biogeographic andt a x o n o m i ci m p l i c at i o n s. J. Plant Res. 112: 3 3 1 – 3 4 1 .

LAGERBERG, T. 1904. Organografiska studier öfverAdoxamoschatellina L. Arkiv for Botanik 3:1–28.

LI, S. H., A N D Z. H. NI N G. 1987. Studies on the fa m i lyA d ox a c e a e.B u l l e t i n of B o t a n i c a l R e s e a rch7 :9 3 – 1 09.

LI A N G, H . 1986. A n a lysis of the karyotype ofTetradoxa omeiensis. Acta Botanica Yunnanica 8:153–156.

— — —. 1997. A comparative study of floral vascu-lature in Adoxaceae. Acta Botanica Yunnanica 19:260–264.

LIANG, H., AND X. ZHANG. 1986. Floral anatomy ofTetradoxa omeiensis. Acta Botanica Yunnanica 8:436–440.

LIN, C. M., Z.-Q. LIU, AND S. D. KING. 1986.Nicotiana ch l o roplast ge n o m e : X. Corre l at i o nbetween the DNA sequences and the isoelectricfocusing pat t e rn of the LS of rubisco. PlantMolecular Biology 6: 81–87.

MADDISON, W. P., AND D. R. MADDISON. 1992.MacClade, Analysis of Phylogeny and CharacterEvolution. Version 3 (computer program). Sinauer,Sunderland, Mass.

MANCHESTER, S. R., AND M. J. DONOGHUE. 1995.Winged fruits of Linnaeeae (Caprifoliaceae) in theTertiary of Western North America: Diplodipeltagen. nov. International Journal of Plant Science156: 709–722.

NEPOMNYASHCHAYA, O. A. 1984. On a new species ofthe genus Adoxa (Adoxaceae) from the far east.Botanicheskii Zhurnal 69: 259–262.

OL M S T E A D, R. G. , B. BR E M E R, K. M. SC OT T, A N D J. D.PA L M E R. 1 9 93. A pars i m o ny analysis of the A s t e ri d a esensu lato based on rbcL sequences. Annals of theMissouri Botanical Garden 80: 700–722.

OLMSTEAD, R. G.,H. J. MICHAELS, K. M. SCOTT, AND

J. D. PALMER. 1992. Monophyly of the Asteridaeand identification of their major lineages inferredfrom DNA sequences of rbcL. Annals of theMissouri Botanical Garden 79: 249–265.

OLMSTEAD, R. G., K.-J. KIM, R. K. JANSEN, AND S. J.WAGSTAFF. 2000. The phylogeny of the Asteridaes e n s u l at o based on ch l o roplast n d hF ge n esequences. Mol. Phylo. Evol. 16: 96–112.

PENG, C.-I., H. TOBE, AND M. TAKAHASHI. 1995.Reproductive morphology and relationships ofTriplosetgia (Dipsacales). Bot. Jahrb. Syst. 116:505–516.

PYCK, N., AND E. SMETS. 2000. A search for the phy-l ogenetic position of the seven-son fl owe r(Heptacodium, Dipsacales): Combining molecularand morphological evidence. Plant. Syst. Evol.225: 185–199.

PYCK, N., P. ROELS, AND E. SMETS. 1999. Tribal rela-tionships in Cap ri fo l i a c e a e : Evidence from acladistic analysis using ndhF sequences. Syst.Geogr. Pl. 69: 145–159.

RA M BAU T, A . 1996. S e - A l : Sequence A l i g n m e n tE d i t o r. Ve rsion 1.0a1 (computer progra m ) .Distributed by the author. Department of Zoology,University of Oxford, Oxford, U.K.

RICE, K. A., M. J. DONOGHUE, AND R. G. OLMSTEAD.1997. Analyzing large data sets: rbcL 500 revis-ited. Systematic Biology 46: 554–563.

ROELS, P., AND E. SMETS. 1994. A comparative floralontogenetical study between Adoxa moschatellinaand Sambucus ebulus. Belgian Journal of Botany127: 157–170.

— — —. 1996. A floral ontogenetic study in theDipsacales. International Journal of Plant Sciences157: 203–218.

STADEN, R. 1996. The Staden sequence analysispackage. Molecular Biotechnology. 5: 233–241.

SWO F F O R D, D. L. 2001. PAUP*. Phy l oge n e t i cAnalysis Using Parsimony (*and Other Methods).Ve rsion 4 (computer program). Sinauer,Sunderland, Mass.

THORNE, R. F. 1983. Proposed new realignments inthe angiosperms. Nordic Jour. Bot. 3: 85–117.

— — —. 1992. An updated classification of the flow-ering plants. Aliso 13: 365–389.

UR BAT S C H, L. E., B. G. BA L DW I N, A N D M. J.DONOGHUE. 2000. Phylogeny of the coneflowers(Heliantheae:Asteraceae) based on nuclear rDNAinternal transcribed spacer sequences. Syst. Bot.25: 539–565.

VE R L AQU E, R . 1977. Rap p o rts entre lesValerianaceae, les Morinaceae et les Dipsacaceae.Bull. Soc. Bot. France 124: 475–482.

WAGENITZ, G., AND B. LAING. 1984. Die Nektariender Dipsacales und ihre systematische Bedeutung.Bot. Jahrb. Syst. 104: 91–113.

WEBERLING, F. 1989. Morphology of flowers andi n fl o rescences. Cambri d ge Unive rsity Pre s s ,Cambridge, U.K.

WILKINSON, A. M. 1948. Floral anatomy and mor-phology of the genus Viburnum. American Journalof Botany 35: 455–465.

— — —. 1949. Floral anatomy and morphology ofTriosteum and of the Caprifoliaceae in general.American Journal of Botany 36: 481–489.

WU, C. Y. 1981. Another new genus of Adoxaceae,with special references on the infrafamiliar evolu-tion and the systematic position of the family. ActaBotanica Yunnanica 3: 383–388.

WU, C. Y., Z. L. WU, AND R. F. HUANG. 1981.Sinadoxa C. Y. Wu,Z. L. Wu et R. F. Huang, genusnovum familiae Adoxacearum. Acta Phytotax. Sin.19: 203–210.

XIANG, Q.-Y., D. E. SOLTIS, P. S. SOLTIS, AND D. R.MO R G A N. 1993. Phy l ogenetic re l ationships ofC o rnu s L. sensu lato and putat ive re l at ive sinferred from rbcL sequence data. Annals of theMissouri Botanical Garden 80: 723–734.

YA KOV L E V, M. S. , A N D G. Y. ZH U KOVA. 1 9 8 0 .Chlorophyll in embryos of angiosperm seeds, areview. Botaniska Notiser 133: 323–336.

2001 DONOGHUE ET AL., DIPSACALES PHYLOGENY 479